Senior Software Engineer

Gradient Robotics; San Francisco (CA); On-site; USD 180,000 - 260,000; Full time

Job summary

Gradient Robotics is hiring a Founding Software Engineer in San Francisco to own the end-to-end data path from AI models to actuators. You will work close to hardware, from kernel and firmware up through controls and perception, building real-time pipelines with millisecond latency.

You’ll shape the software stack to keep vision inference, control loops, and actuation on a single clock, ensuring determinism and high performance for the autonomous robotics platform.

Qualifications

• 5+ years of software engineering near hardware (drivers, kernels, embedded, robotics).
• Proficient in Rust, Python or C++, with strong multithreading skills.
• Experience with real-time or deterministic systems in robotics.
• Familiarity with Bazel, Nix, and CI/CD pipelines.
• Experience with Linux kernel hacking and embedded platforms.

Responsibilities

• Ship performance-critical code to real robots daily.
• Own the full data flow: camera frames, perception, planning, and actuators.
• Maintain deterministic real-time control loops alongside vision inference.
• Develop low-latency perception data paths for high-bandwidth streams.
• Reduce latency across the stack and increase system reliability.
• Build tooling and instrumentation to observe and debug real-time behavior.

Skills

Rust

Python

C++

Multithreading

Embedded systems

Tools

Bazel

Nix

CI/CD

Linux kernel hacking

Embedded Linux

DPDK

SPDK

io_uring

eBPF

ftrace

flamegraphs

Job description

About the Company

The datacenter buildout is the largest industrial project in human history. Gradient builds the autonomous robots that make it possible.

Partnered with the world’s largest AI infrastructure companies and backed by the industry’s best investors, we move fast and build full-stack systems that matter.

About the role

We’re looking for a Founding Software Engineer to own how Gradient Robotics moves data from AI models to actuators. You’ll work close to the hardware, from kernel and firmware up through the controls and perception layers, owning real-time pipelines that move hundreds of megabytes at single-digit millisecond latency. Vision inference, control loops, and actuation all live on the same clock, and it’s your job to keep them there. The goal: bring visibility and determinism into the end-to-end inference pipeline so the ML model is the only stochastic component.

Responsibilities

• Ship performance-critical code to real robots daily

• Own the full data flow: camera frames in, perception and planning in the middle, control commands and actuator feedback out, and every system that carries them

• Own the real-time control loop: keep high-rate control running deterministically alongside vision inference, and make sure a slow frame never becomes a late motor command

• Build the perception data path: move high-bandwidth camera streams into the CV and ML models with minimal latency and zero silent drops

• Push down latency and tighten the stack to make the system faster, more deterministic, and more reliable

• Bring visibility into the pipeline: build the tooling and instrumentation that make real-time behavior across controls and vision observable and debuggable

Minimum Qualifications

• 5+ years of software engineering experience building production systems close to hardware (drivers, kernels, embedded, robotics, or similar)

• Software: Rust, Python, C++, operating systems, and multithreading

• Infrastructure: Bazel, Nix, HIL testing, and CI/CD

• Firmware and platform: Linux kernel hacking and embedded systems

• Debugged timing and concurrency issues in the wild

• Owned messy, ambiguous problems and turned them into robust software

• Able to work on-site in San Francisco 5 days/week (6 days/week if needed during crunch time), embedded in the team

Preferred Qualifications

• Experience with real-time control systems: motor control, high-rate control loops, or robot controllers

• Computer vision pipeline experience: camera drivers, image transport, GPU inference, or sensor synchronization

• Tracing and profiling experience: ftrace, flamegraphs, eBPF

• Linux I/O experience: DPDK, SPDK, io_uring

• Comfort using AI tools to multiply your own output